detection of fatty tissue in HE images

[Marien Koen]

Dear ImageJ mailing list


Currently I'm trying to write an algorithm that can detect fatty tissue in an image of HE stained tissue (examples: https://www.dropbox.com/sh/40unigstsq795tk/AADpFIvIbudvDCJqdNxINa1_a?dl=0).
Making a difference between holes in the tissue and the fatty is not easy.

Can somebody help me with a succesful workflow? With choosing the right filters? Or if you know good literature about this?


Thanks
Koen


ir. Koen Marien

Imaging Scientist, PhD student

<file://hgx-storage2/General/Templates/HGX%20logooke.jpg>

Lab building, 1st floor (campus ZNA Middelheim)

Lindendreef 1, 2020 Antwerpen (Belgium)        Phone: [phone]

https://www.uantwerp.be/en/staff/koen-marien/

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Volker Baecker]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512

Hello,
I did something similar, see

http://dev.mri.cnrs.fr/projects/imagej-macros/wiki/Adipocytes_Tool

You will need to find the right parameters and maybe adapt the macros
to your needs.

Alternatively there is

Osman OS, Selway JL, K?pczy?ska MA, Stocker CJ, O’Dowd JF, Cawthorne
MA, et al. A novel automated image analysis method for accurate
adipocyte quantification. Adipocyte. 2013 Jul 1;2(3):0–1.

using MATLAB

Best regards,
Volker Baecker

Marien Koen wrote:
> Dear ImageJ mailing list
>
>
> Currently I'm trying to write an algorithm that can detect fatty
> tissue in an image of HE stained tissue (examples:
> https://www.dropbox.com/sh/40unigstsq795tk/AADpFIvIbudvDCJqdNxINa1_a?dl=0).
>
>
Making a difference between holes in the tissue and the fatty is not easy.

>
> Can somebody help me with a succesful workflow? With choosing the
> right filters? Or if you know good literature about this?
>
>
> Thanks Koen
>
>
> ir. Koen Marien
>
> Imaging Scientist, PhD student
>
> <file://hgx-storage2/General/Templates/HGX%20logooke.jpg>
>
> Lab building, 1st floor (campus ZNA Middelheim)
>
> Lindendreef 1, 2020 Antwerpen (Belgium)        Phone: [phone]
> 19 17
>
> https://www.uantwerp.be/en/staff/koen-marien/
>
> -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

-----BEGIN PGP SIGNATURE-----

iQEcBAEBCgAGBQJUAD7gAAoJENcVU956o58NbiMIAJWCBRMYPYFDnAkUxKMp2Gms
kiNSJSWSxZ++Slf8DBCRmT1ijmrj6Fg7vQjdA204x703FD2g5kjzlasQYYfcuoJ6
u+HLlG0wvoQq0JkZOYIPVcwXZTlevCJAZaK+NQCHNuPCZ34RsGbTIv6bF412X7aO
NLJ4Hy2rPnv2LSg0xJ+/q40MxptdnvYz2e316kpHXOygoxdp8V6rWKD4PFvmzs8g
iGzILsY3HDl9ivIS/fWiWPE7hGb32uImBtO4ocsAcyg8IlJnnu9tRNdfkMXGH0M9
5t9+keNKxEC7yc2KRKW68gfaFdea78ShzWP/H2hfb2eu15twQW9SqA5t4RBxCso=
=zq0H
-----END PGP SIGNATURE-----

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Lendl,Thomas]

Hi Marien,

If you are just interested in tissue sizes and not in single cells, this could be a possible approach for you:
1.Use the green Channel for segmentation, make a copy of it.
2.Define tissue versus Background by using the FeatureJ Edges filter and thresholding on that.
3. Create a mask from the thresholded Image, reshape your region via binary operations (dilation and closing, fill holes) to cover the whole tissue.
4. Create a selection on the binary image, measure the size, transfer it to your original green channel.
5. Clear outside so that you get a black background. Smooth once.
6. Threshold again on the brighter fatty tissue, make a selection and measure the size. If necessary you can create a binary mask again to reshape your objects before measuring.

This should be doable to automate with a macro and run on a batch of images, at least if they are similar enough.

Just tell me if this worked out for you.
Best,
Thomas


-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Volker Baecker
Sent: Friday, August 29, 2014 10:51 AM
To: [hidden email]
Subject: Re: detection of fatty tissue in HE images

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512

Hello,
I did something similar, see

http://dev.mri.cnrs.fr/projects/imagej-macros/wiki/Adipocytes_Tool

You will need to find the right parameters and maybe adapt the macros to your needs.

Alternatively there is

Osman OS, Selway JL, K?pczy?ska MA, Stocker CJ, O'Dowd JF, Cawthorne MA, et al. A novel automated image analysis method for accurate adipocyte quantification. Adipocyte. 2013 Jul 1;2(3):0-1.

using MATLAB

Best regards,
Volker Baecker

Marien Koen wrote:
> Dear ImageJ mailing list
>
>
> Currently I'm trying to write an algorithm that can detect fatty
> tissue in an image of HE stained tissue (examples:
> https://www.dropbox.com/sh/40unigstsq795tk/AADpFIvIbudvDCJqdNxINa1_a?dl=0).
>
>
Making a difference between holes in the tissue and the fatty is not easy.

>
> Can somebody help me with a succesful workflow? With choosing the
> right filters? Or if you know good literature about this?
>
>
> Thanks Koen
>
>
> ir. Koen Marien
>
> Imaging Scientist, PhD student
>
> <file://hgx-storage2/General/Templates/HGX%20logooke.jpg>
>
> Lab building, 1st floor (campus ZNA Middelheim)
>
> Lindendreef 1, 2020 Antwerpen (Belgium)        Phone: [phone]
> 19 17
>
> https://www.uantwerp.be/en/staff/koen-marien/
>
> -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

-----BEGIN PGP SIGNATURE-----

iQEcBAEBCgAGBQJUAD7gAAoJENcVU956o58NbiMIAJWCBRMYPYFDnAkUxKMp2Gms
kiNSJSWSxZ++Slf8DBCRmT1ijmrj6Fg7vQjdA204x703FD2g5kjzlasQYYfcuoJ6
u+HLlG0wvoQq0JkZOYIPVcwXZTlevCJAZaK+NQCHNuPCZ34RsGbTIv6bF412X7aO
NLJ4Hy2rPnv2LSg0xJ+/q40MxptdnvYz2e316kpHXOygoxdp8V6rWKD4PFvmzs8g
iGzILsY3HDl9ivIS/fWiWPE7hGb32uImBtO4ocsAcyg8IlJnnu9tRNdfkMXGH0M9
5t9+keNKxEC7yc2KRKW68gfaFdea78ShzWP/H2hfb2eu15twQW9SqA5t4RBxCso=
=zq0H
-----END PGP SIGNATURE-----

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html